zedstat 0.0.151__tar.gz → 0.0.153__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- {zedstat-0.0.151/zedstat.egg-info → zedstat-0.0.153}/PKG-INFO +2 -2
- zedstat-0.0.153/version.py +1 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/zedstat.py +172 -0
- {zedstat-0.0.151 → zedstat-0.0.153/zedstat.egg-info}/PKG-INFO +2 -2
- zedstat-0.0.151/version.py +0 -1
- {zedstat-0.0.151 → zedstat-0.0.153}/LICENSE +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/MANIFEST.in +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/README.rst +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/setup.cfg +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/setup.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/.ipynb_checkpoints/Untitled-Copy1-checkpoint.ipynb +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/.ipynb_checkpoints/Untitled-checkpoint.ipynb +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/.ipynb_checkpoints/example3-checkpoint.ipynb +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/__init__.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/calibration.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/ci.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/example_combine.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/mwauc.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/rrci.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/textable.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat/zcombine.py +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat.egg-info/SOURCES.txt +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat.egg-info/dependency_links.txt +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat.egg-info/requires.txt +0 -0
- {zedstat-0.0.151 → zedstat-0.0.153}/zedstat.egg-info/top_level.txt +0 -0
|
@@ -1,9 +1,9 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: zedstat
|
|
3
|
-
Version: 0.0.
|
|
3
|
+
Version: 0.0.153
|
|
4
4
|
Summary: Statistics tools for ML models and deployment
|
|
5
5
|
Home-page: https://github.com/zeroknowledgediscovery/zedstat
|
|
6
|
-
Download-URL: https://github.com/zeroknowledgediscovery/zedstat/archive/0.0.
|
|
6
|
+
Download-URL: https://github.com/zeroknowledgediscovery/zedstat/archive/0.0.153.tar.gz
|
|
7
7
|
Author: zed.uchicago.edu
|
|
8
8
|
Author-email: ishanu@uchicago.edu
|
|
9
9
|
License: LICENSE
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
__version__ = '0.0.153'
|
|
@@ -1746,3 +1746,175 @@ try:
|
|
|
1746
1746
|
__all__.append('lr_nondominated_frontier')
|
|
1747
1747
|
except Exception:
|
|
1748
1748
|
pass
|
|
1749
|
+
|
|
1750
|
+
|
|
1751
|
+
# =============================================================================
|
|
1752
|
+
# NaN-safe final output accessor
|
|
1753
|
+
# =============================================================================
|
|
1754
|
+
# This accessor keeps the old behavior for zt.get(): nominal table only, no
|
|
1755
|
+
# interpolation. It adds two explicit final-output modes:
|
|
1756
|
+
# zt.get(bounds=True, interpolate=True)
|
|
1757
|
+
# returns nominal + upper/lower bounds and interpolates display gaps;
|
|
1758
|
+
# zt.get(0)
|
|
1759
|
+
# shorthand for the same NaN-safe final table. The 0 is not used as a
|
|
1760
|
+
# numerical fill value; residual all-NaN columns remain NaN unless fillna
|
|
1761
|
+
# is supplied explicitly.
|
|
1762
|
+
|
|
1763
|
+
_PROBABILITY_BASE_COLUMNS = {'tpr', 'ppv', 'acc', 'npv'}
|
|
1764
|
+
|
|
1765
|
+
|
|
1766
|
+
def _base_metric_name(col):
|
|
1767
|
+
col = str(col)
|
|
1768
|
+
for suffix in ('_upper', '_lower'):
|
|
1769
|
+
if col.endswith(suffix):
|
|
1770
|
+
return col[:-len(suffix)]
|
|
1771
|
+
return col
|
|
1772
|
+
|
|
1773
|
+
|
|
1774
|
+
def _clip_probability_like_columns(frame):
|
|
1775
|
+
out = frame.copy()
|
|
1776
|
+
for col in out.columns:
|
|
1777
|
+
if _base_metric_name(col) in _PROBABILITY_BASE_COLUMNS:
|
|
1778
|
+
vals = pd.to_numeric(out[col], errors='coerce').to_numpy(dtype=float)
|
|
1779
|
+
out[col] = np.clip(vals, 0.0, 1.0)
|
|
1780
|
+
return out
|
|
1781
|
+
|
|
1782
|
+
|
|
1783
|
+
def _interpolate_numeric_frame_for_display(frame, interpolate=True, fillna=None, limit_direction='both'):
|
|
1784
|
+
"""Interpolate numeric display gaps in a final performance table.
|
|
1785
|
+
|
|
1786
|
+
This is intentionally a final accessor/post-processing step. It does not
|
|
1787
|
+
change the stored analytic bounds unless the caller assigns the returned
|
|
1788
|
+
dataframe. The main use is to prevent visualization breaks from NaNs that
|
|
1789
|
+
arise after endpoint masking, sparse empirical ROC rows, or frontier masking.
|
|
1790
|
+
"""
|
|
1791
|
+
out = frame.copy().replace([np.inf, -np.inf], np.nan)
|
|
1792
|
+
|
|
1793
|
+
if interpolate:
|
|
1794
|
+
numeric_cols = [c for c in out.columns if pd.api.types.is_numeric_dtype(out[c])]
|
|
1795
|
+
if numeric_cols:
|
|
1796
|
+
idx_numeric = np.issubdtype(out.index.dtype, np.number)
|
|
1797
|
+
method = 'index' if idx_numeric else 'linear'
|
|
1798
|
+
out[numeric_cols] = out[numeric_cols].interpolate(
|
|
1799
|
+
method=method,
|
|
1800
|
+
limit_direction=limit_direction,
|
|
1801
|
+
axis=0,
|
|
1802
|
+
)
|
|
1803
|
+
|
|
1804
|
+
if fillna is not None:
|
|
1805
|
+
out = out.fillna(fillna)
|
|
1806
|
+
|
|
1807
|
+
out = _clip_probability_like_columns(out)
|
|
1808
|
+
return out
|
|
1809
|
+
|
|
1810
|
+
|
|
1811
|
+
def _enforce_joined_bounds_for_display(frame):
|
|
1812
|
+
"""Ensure joined *_lower and *_upper columns bracket nominal columns."""
|
|
1813
|
+
out = frame.copy()
|
|
1814
|
+
for base in ['tpr', 'ppv', 'acc', 'npv', 'LR+', 'LR-']:
|
|
1815
|
+
lo_col = f'{base}_lower'
|
|
1816
|
+
hi_col = f'{base}_upper'
|
|
1817
|
+
if base not in out.columns or lo_col not in out.columns or hi_col not in out.columns:
|
|
1818
|
+
continue
|
|
1819
|
+
|
|
1820
|
+
n = pd.to_numeric(out[base], errors='coerce').to_numpy(dtype=float)
|
|
1821
|
+
lo = pd.to_numeric(out[lo_col], errors='coerce').to_numpy(dtype=float)
|
|
1822
|
+
hi = pd.to_numeric(out[hi_col], errors='coerce').to_numpy(dtype=float)
|
|
1823
|
+
finite = np.isfinite(n) & np.isfinite(lo) & np.isfinite(hi)
|
|
1824
|
+
if not np.any(finite):
|
|
1825
|
+
continue
|
|
1826
|
+
|
|
1827
|
+
lo2 = lo.copy()
|
|
1828
|
+
hi2 = hi.copy()
|
|
1829
|
+
lo2[finite] = np.minimum.reduce([lo[finite], hi[finite], n[finite]])
|
|
1830
|
+
hi2[finite] = np.maximum.reduce([lo[finite], hi[finite], n[finite]])
|
|
1831
|
+
|
|
1832
|
+
if base in _PROBABILITY_BASE_COLUMNS:
|
|
1833
|
+
lo2 = np.clip(lo2, 0.0, 1.0)
|
|
1834
|
+
hi2 = np.clip(hi2, 0.0, 1.0)
|
|
1835
|
+
|
|
1836
|
+
out[lo_col] = lo2
|
|
1837
|
+
out[hi_col] = hi2
|
|
1838
|
+
return out
|
|
1839
|
+
|
|
1840
|
+
|
|
1841
|
+
def _processRoc_get_nan_safe(
|
|
1842
|
+
self,
|
|
1843
|
+
interpolate=False,
|
|
1844
|
+
bounds=False,
|
|
1845
|
+
fillna=None,
|
|
1846
|
+
limit_direction='both',
|
|
1847
|
+
enforce_bounds=True,
|
|
1848
|
+
):
|
|
1849
|
+
"""Return the processed ROC table, optionally as a NaN-safe final table.
|
|
1850
|
+
|
|
1851
|
+
Parameters
|
|
1852
|
+
----------
|
|
1853
|
+
interpolate : bool or numeric, default False
|
|
1854
|
+
False preserves the historical zt.get() behavior. True interpolates
|
|
1855
|
+
numeric gaps in the returned display table. Passing 0 is supported as a
|
|
1856
|
+
compact shorthand for zt.get(bounds=True, interpolate=True).
|
|
1857
|
+
bounds : bool, default False
|
|
1858
|
+
If True, join self.df_lim['U'] and self.df_lim['L'] using _upper and
|
|
1859
|
+
_lower suffixes before optional interpolation.
|
|
1860
|
+
fillna : scalar or None, default None
|
|
1861
|
+
Optional final fill for columns that remain NaN after interpolation.
|
|
1862
|
+
Leave as None to avoid inventing values for all-NaN columns.
|
|
1863
|
+
limit_direction : str, default 'both'
|
|
1864
|
+
Passed to pandas interpolate. 'both' fills endpoint display gaps using
|
|
1865
|
+
nearest finite values.
|
|
1866
|
+
enforce_bounds : bool, default True
|
|
1867
|
+
After interpolation, force lower <= nominal <= upper for joined bounds.
|
|
1868
|
+
"""
|
|
1869
|
+
numeric_shorthand = isinstance(interpolate, (int, float, np.integer, np.floating)) and not isinstance(interpolate, (bool, np.bool_))
|
|
1870
|
+
if numeric_shorthand:
|
|
1871
|
+
# zt.get(0) means: return the final joined, interpolated display table.
|
|
1872
|
+
# The zero is not treated as a fill value because LR+ at FPR=0 should not
|
|
1873
|
+
# be replaced by zero. Use fillna=0 explicitly if that is actually wanted.
|
|
1874
|
+
if float(interpolate) != 0.0 and fillna is None:
|
|
1875
|
+
fillna = float(interpolate)
|
|
1876
|
+
interpolate = True
|
|
1877
|
+
bounds = True
|
|
1878
|
+
|
|
1879
|
+
out = self.df.copy()
|
|
1880
|
+
|
|
1881
|
+
if bounds:
|
|
1882
|
+
if 'U' not in self.df_lim or 'L' not in self.df_lim:
|
|
1883
|
+
raise ValueError("Bounds requested, but getBounds() has not been run yet.")
|
|
1884
|
+
out = (
|
|
1885
|
+
out
|
|
1886
|
+
.join(self.df_lim['U'], rsuffix='_upper')
|
|
1887
|
+
.join(self.df_lim['L'], rsuffix='_lower')
|
|
1888
|
+
)
|
|
1889
|
+
|
|
1890
|
+
if interpolate or fillna is not None:
|
|
1891
|
+
out = _interpolate_numeric_frame_for_display(
|
|
1892
|
+
out,
|
|
1893
|
+
interpolate=bool(interpolate),
|
|
1894
|
+
fillna=fillna,
|
|
1895
|
+
limit_direction=limit_direction,
|
|
1896
|
+
)
|
|
1897
|
+
if bounds and enforce_bounds:
|
|
1898
|
+
out = _enforce_joined_bounds_for_display(out)
|
|
1899
|
+
|
|
1900
|
+
return out.copy()
|
|
1901
|
+
|
|
1902
|
+
|
|
1903
|
+
def _processRoc_get_full(self, interpolate=True, fillna=None, limit_direction='both', enforce_bounds=True):
|
|
1904
|
+
"""Convenience wrapper for the final nominal + bounds table."""
|
|
1905
|
+
return self.get(
|
|
1906
|
+
bounds=True,
|
|
1907
|
+
interpolate=interpolate,
|
|
1908
|
+
fillna=fillna,
|
|
1909
|
+
limit_direction=limit_direction,
|
|
1910
|
+
enforce_bounds=enforce_bounds,
|
|
1911
|
+
)
|
|
1912
|
+
|
|
1913
|
+
|
|
1914
|
+
processRoc.get = _processRoc_get_nan_safe
|
|
1915
|
+
processRoc.get_full = _processRoc_get_full
|
|
1916
|
+
|
|
1917
|
+
try:
|
|
1918
|
+
__all__.append('lr_nondominated_frontier')
|
|
1919
|
+
except Exception:
|
|
1920
|
+
pass
|
|
@@ -1,9 +1,9 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: zedstat
|
|
3
|
-
Version: 0.0.
|
|
3
|
+
Version: 0.0.153
|
|
4
4
|
Summary: Statistics tools for ML models and deployment
|
|
5
5
|
Home-page: https://github.com/zeroknowledgediscovery/zedstat
|
|
6
|
-
Download-URL: https://github.com/zeroknowledgediscovery/zedstat/archive/0.0.
|
|
6
|
+
Download-URL: https://github.com/zeroknowledgediscovery/zedstat/archive/0.0.153.tar.gz
|
|
7
7
|
Author: zed.uchicago.edu
|
|
8
8
|
Author-email: ishanu@uchicago.edu
|
|
9
9
|
License: LICENSE
|
zedstat-0.0.151/version.py
DELETED
|
@@ -1 +0,0 @@
|
|
|
1
|
-
__version__ = '0.0.151'
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
{zedstat-0.0.151 → zedstat-0.0.153}/zedstat/.ipynb_checkpoints/Untitled-Copy1-checkpoint.ipynb
RENAMED
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|
|
File without changes
|